GridDB Reborn!

A full nine years have passed since GridDB was released in 2013, and the software structure has become more complex as a result of the various enhancements that have been incorporated during that time. Therefore, in GridDB V5, the software structure has been renewed to improve maintainability and provide a foundation for rapid future enhancements. GridDB has been reborn so that GridDB can be used for many years to come. 

In the past, GridDB developed Event-Driven Engine and Autonomous Data Distribution Algorithm (ADDA) to support Big Data and IoT systems that require fast, scalable and reliable data store. However, in recent years, IoT data and its utilization have diversified immensely resulting in the need to handle different types of data using different data models accordingly. The current solutions are to utilize multiple Database Management Systems (DBMS) or to store all data into a single data model, which introduces a variety of problems such as system complexity, high operational costs, and a drop in performance.

GridDB 5.0 Enterprise Edition comes with a revamped architecture that features pluggable data store where multiple data model can be managed in
a single DBMS. In addition to the current data store optimized for high-frequency high-volume data ingestion, other data stores for executing complex analyses at high speed and for storing text such as logs can be incorporated.

Recently, in addition to the value provided by IoT systems in storing and visualizing large amounts of sensor data, there has been a growing need to
utilize the data by performing complex analyses to gain new business insights. The ability to store large volumes of high frequency data and perform complex analyses at high speed are conflicting requirements for a DBMS. 

The pluggable data store function makes it possible to integrate multiple data stores optimized for specific workload into a single DBMS. Rather
than using multiple DBMSs, integrated processing can be executed in a single DBMS and thus avoiding increased system complexity, and higher construction and operational costs.

A data store that is able to perform complex analyses at high speed and another optimized for storing text data will be provided sequentially in the future.

In GridDB 5.0, the unique checkpoint algorithm based on HCAL (Highly Efficient Checkpoint Algorithm) results in a reduced amount of log written to files during checkpoint and lowers the disk I/O load. Database performance can be improved by reducing the load to systems that frequently perform data ingestion and update.

In addition, unique blocks can now be assigned to each table to speed up table scan and deletion. This is particularly effective for data analysis queries where table scan is frequently utilized. Table deletion performance is also improved as tables to be deleted can be specified.

According to a database benchmark test (TPC-H), these performance enhancements have resulted in improvements in the region of 17% to 46% (26% on average).